Professional Documents
Culture Documents
CEES P. MIDDENDORP
Niemoller and Van der Eijk (1986a, 1987) have recently shown that left-right self-
identification is an important determinant of the vote-if the parties are ordered
according to the mean left-right scores assigned to them by the electorate. They
have also maintained that this is not a trivial relationship (NiemMer and Van der
Eijk, 1986a, see also Van Deth, 1986). If one orders the parties according to the
mean scores on materialism or post-materialism of their voters (see IngIehart, 1977)
the predictive power of left-right self-identification remains strong whereas the
effect of materialism or post-materialism remains relatively weak. In this context they
consider (post)materialism an expression of left-right self-identification, and show
that this latter variable is hardly determined at all by social characteristics such as
subjective social class (LISREL-coefficient: 0.29) educational level (-0.24) and age
(0.14). (In the 1987 study, religion proves to be a relatively strong determinant of
The ideological space in the Netherlands has been established on the basis of an
analysis of the progressive-conservative domain of ideological controversy.z This
domain has been defined by means of the construction of an ideal type conceptual
model. This model has been ‘analysed’ in terms of underlying values (equality and
freedom) and their application to the socioeconomic and the politico-cultural
realms, respectively. Subsequently, attitude scales were constructed measuring each
of the values involved and their applications to the two fields.3
In a series of studies since 1970, at each point in time (1975, 1980, 1985) a stable
two-dimensional structure was found in the Dutch electorate; each dimension
encompassing about 10 attitude scales.’
The first dimension in the socio-ecorionzic feft-right dimension, built up from
attitudes regarding, roughly, equality in socio-economic terms, as well as derived
attitudes on government welfare state policies, that is, attitudes regarding social
welfare policies, income policies, educational policies, policies on taxation,
CEE~P. MIDDENDOW 251
AUTHORITARIAN
GPV 0
(Calvinist
SGP q Fundamentalists)
:’
(3)
Extreme rightwing
4--___ CDA
---_
LEFT-WING ---m
(1) RIGHT-WING
I I
I
I
---- I
- - ._
-iWD
(Liberals)
Extreme
leftwing
0 0
Communists Democrats 66
(CPN)
0
Pacifist Socialists
LIBERTARIAN
(PSP)
FIG. 1. Mean scores for party supporters and left-right self-identifiers on dimensions in the
ideological space; the Dutch electorate, 1985
n = Major party: PvdA=Iabour Party; CDA=Christian-Democratic Appeal, VVD=L.iberals
q = Minor party. Angle between axes indicates minor correlation of 0.19.
On the socioeconomic left-right dimension, the liberal VVD supporters are more
right wing than the Christian Democrats (CDA) but on libertarianism-authoritarian-
ism, the latter are more authoritarian than the liberals. Furthermore, the Democrats
are more libertarian than the Labour Party (PvdA) supporters, but the latter are
more left wing socioeconomically. It can also be seen in Fig. 1 that the mean scores
on both ideological dimensions for those considering themselves either (extremely)
left wing or right wing can be plotted in between the axes indicating socio-
economic left-right and politico-cultural libertarian-authoritarian ideology. The
ordering of the mean scores for the various political party supporters along this line
is very similar to the ordering of the supporters along the socioeconomic left-right
dimension, but different from the ordering of the party supporters along the liber-
tarian-authoritarian dimension.
Left-right self-identification is as strongly associated to socioeconomic left-right
ideology as to libertarian-authoritarian ideology (Pearson r’s are 0.49 and 0.47
respectively; compare Inglehart, 1984:31-45; Flanagan, 1982:435). The conclusion
must be that this self-identification lacks an unambiguous content validity. At least
we have to conclude that the content-validity of left-right self-identification is
dualistic, rather than unequivocal.
Obviously, in explaining the vote, the political parties can be ordered in two differ-
ent manners (see Fig. 1). Along one ideological dimension, the socioeconomic
left-right vote can be measured; along the other, the libertarian-authoritarian vote.’
Assuming for the moment that left-right self-identification is dependent upon (a
‘summary’ variable springing from) both ideological stands, the vote can be
predicted in two ways, as is shown in Fig. 2. From this figure it is clear that it is
highly dependent upon the way the various parties are ordered which variables are
major determinants of the vote.
For the socioeconomic left-right vote, left-right self-identification has the
strongest direct and total effect. The effects of socioeconomic left-right ideology,
class and religion (involvement; church attendance) are clearly weaker.
The direct and total effects of libertarianism-authoritarianism on the left-right
vote are rather weak (0.09 and 0.25 respectively). The effects of social class and
religion are largely interpreted by the ideological dimensions and left-right self-
identification; the direct effects of these variables on the left-right vote are weak
(0.10 and 0.11 respectively). In 1985, almost 60 per cent of the socioeconomic
left-right vote can be explained by the model.
The picture for the libertarian-authoritarian vote is quite different. Here, left-right
self-identification has the strongest direct effect but religious involvement has the
strongest total effect, whereas libertarian-authoritarian ideology also has a relatively
strong total effect, as has socioeconomic left-right ideology. The effects of social class
and income have become rather weak. Left-right self-identification has also a relatively
strong effect on the libertarian-authoritarian vote because the two voting patterns are
strongly correlated (0.79). The strong total religion effect is brought about by many
indirect effects of the religious factor. The effect of religion on the libertarian-authori-
tarian vote is stronger than the effect of social class on the left-right vote, but on the
other hand socioeconomic left-right ideology has a stronger effect on the left-right
vote than libertarianism-authoritarianism has on the libertarian-authoritarian vote.
Left-right
Ideology
n
vote
vote 1
7 11(; !6)
PY/
Religion / /
Notes:
Fit for left-right vote: Chi-2=14.14 (Df=lO), p=.16; Explained variance: 59%.
Fit for Libertarian-authoritarian vote: Chi-2=17.25 (Df=ll), p=.lO; Explained variance=Sl%.
Total effects on the vote: LR Self-identif. LR Ideology Lib-auth. Subj.class Income Religion
FIG. 2. Models explaining the left-right and libertarian-authoritarian vote, with left-right self-identification as a result of
ideological stands
CEES P. MIDDENDORP 255
Alternative Models
Niemijller and Van der Eijk (1986a, 1987) consider (post)materialism an expression
of left-right self-identification. Theoretically, this position seems weak since it
requires the assumption that complex value orientations are somehow caused by
subjective self-placements (see above). If this is nevertheless done (i.e., by means
of changing the order of the position of the two ideologies and left-right self-identi-
fication in the causal chain represented in the model in Fig. 2) then the total effect
of left-right self-identification on the vote naturally becomes stronger due to
additional Indirect effects through the two ideologies. Even in this case, however,
left-right self-identification is not the strongest determinant of the libertarian-
authoritarian vote: the religion effect is evenly strong.
Left-right self-identification itself is only moderately determined by social charac-
teristics (including religion: 17 per cent of the variance is explained) whereas in
the other models almost 40 per cent could be explained by this variable because
of the additional effects of the ideological dimensions. It seems that the Niemoller
Van der Eijk position regarding the overwhelming effect of left-right self-identifi-
cation on ‘the’ vote is only supported when (a) the parties are ordered along the
left-right dimension, and (b) left-right self-identification is seen as a determinant
of ideological stands rather than as a result of them. The latter position seems
theoretically preferable, however. I will now consider their position regarding the
role of (post)materialism which has been challenged by Van Deth and Geurts
(1989).
The Niemiiller and Van der Eijk (1987) model has been retested in the following
manner: (1) reversing the positions of left-right self-identification and (post)materi-
alism and (2) with the two alternative orderings of the parties: according to the
left-right dimension and according to the libertarian-authoritarian dimension.’ The
results are presented in Fig. 3.
It can be seen there that (a) the socio-economic left-right vote is most strongly
determined by left-right self-identification; (b) the direct effects of (post)material-
ism on the left-right vote as well as on the religious vote are weak and the total
effects are only moderate (the latter coefficients are 0.29 and 0.24 respectively);
(c) the socio-economic left-right vote is more strongly determined by religion (0.46)
Notes:
a) Data are from the Dutch National Election Study, 1982. Model fits at .03 level.
b) The Libertarian-authoritarian vote is virtually identical to the religious vote, in which the parties are ordered according the
frequency of church attendance of their supporters. Coefficients for the religious vote are in brackets.
Total effects on: of: Left-R Self-identif. Post-materialism Subj .class Income Religion (Church-att.)
Left-right vote .63 -.29 .35 .19 .46
Religious vote .48 -.24 .29 .Ol .62
See note 7 for the orderings of the political parties.
FIG. 3. Left-right self-identification, (post)materialism, and the left-right and libertarian-authoritarian vote (i.e. the
religious votey
_ _ _ __
CEES P. MIDDENDORP 257
than by social class (0.35) and (d) religion has the strongest total effect on the
religious (libertarian-authoritarian) vote (0.62 as compared with 0.48 for left-right
self-identification; see also Broos, 1987).
Alongside the fact that (post)materialism is of minor importance as a determinant
of the vote (which confiis Niemoller and Van der Eijk; 1986a, 1987, and is in
contradistinction to Inglehart (1990) Inglehart and Rabier (1986) and contradicts
Van Deth and Geurts (1989)) it is again shown that it depends on the ordering of
the parties which variable predicts the vote best. If the parties are ordered along
the socio-economic left-right dimension, left-right self-identification is the best
predictor, if, however, they are ordered along the religious or libertarian-authori-
tarian dimension, then religion is the best predictor. Although left-right self-identi-
fication may thus be the best predictor of the vote in particular circumstances, it
is theoretically less attractive because, as was shown above, it lacks a clear and
unambiguous content validity (see also Van Deth, 1986).
The effect of (post)materialism-measured according to Inglehart’s (1977) 12-
item battery-on the vote is comparable to the effect of libertarian-authoritarian
ideology on the vote (compare Fig. 2 and 3). I agree with Flanagan (1982, 1987)
and Knutsen (1986, 1988, 1989, 1990) that the libertarian-authoritarian elements in
(post)materialism are most relevant politically, that is, for the vote. The ‘economic’
aspects of materialism (inflation, growth, stable economy)-apart from the ‘authori-
tarian’ ones (order, strong army, fight crime) seem to be less relevant politically
and ‘ideologically’ and so are the ‘non-libertarian’ aspects of post-materialism
(friendlier and less impersonal society; society in which ideas are more important
than money). The political ‘kernel’ of post-materialism seems to be its libertarian-
authoritarian aspects!
In this study it has been confiied (see Niemoller and Van der Eijk, 1987) that
left-right self-identification is indeed the best predictor of the left-right vote-that
is, if the parties are ordered from left to right on the basis of left-right scores
assigned to them by voters or according to the mean left-right positions of their
supporters on the socioeconomic ideological left-right dimension. However, the
question about the substantial meaning of the terms left and right has not been
sufficiently dealt with in the previous work of Van der Eijk and Niemoller (1983).
In this study it has been shown that there is a stable two-dimensional ideological
space in the Dutch electorate since 1970 (see Middendorp, 1978, 1991; in this
article only 1985data are presented). One dimension-built up from many attitude
scales (see Table 1 above, and for details Middendorp, 1978; 1991)-is the socio-
economic left-right dimension; the other dimension, virtually independent from the
former one, is called politico-cultural libertarianism-authoritarianism. Both dimen-
sions are equally strongly associated (in 1985) to people’s self-identification in terms
of left wing or right wing. Thus, the conclusion is inevitable that left-right self-
identification lacks unambiguous content validity. It was also shown in this study
that the predictive power of left-right self-identification is dependent on the
manner in which the political parties are ordered. Although Niemoller and Van der
Eijk have shown that left-right self-identification remains the strongest predictor of
the vote even when the parties are ordered along the (post)materialism dimension,
I have shown above that in case the parties are ordered along the other major
258 Left-Right Self-identification
Notes
1. Field work for the 1975, 1980 and 1985studies was carried out by the Netherlands Insti-
tute for Public Opinion and Market Research (NIP0 BV), Amsterdam. The project was
(since 1975) carried out in colla~mtion with the Social and Cultural Planning Bureau
(SCP) of the Ministry of Welfare, Health and Culture (WVC). The series of studies was
sponsored by the Netherlands Organization for the Advancement of Research (NWO;
previously ZWO), The Hague (except the 1975study). For a ‘complete’ report on the
series of studies, see Middendo~ (1991).
2. For the actual approach followed, see Middendorp (1978 chapters 4 and 5, and 1991,
chapter 3). See Middendorp (199la) for a summary of this approach, in general method-
ological terms.
3. All attitude scales used are constructed by means of Mokken’s (1971) procedure (see
below, Table 1 and again Middendorp, 19778, 1991, App.4).
4. A full report on the series of studies has been published in Middendorp (1991). Data
are available on request from the Steinmetz Archives, SWIDOC, Royal Netherlands
Academy of Arts and Science, Amsterdam.
5. At a more abstract level, the progressive conservative ideal-type model was opera-
tionalized ‘directly’ and empirically assessed in the Dutch electorate. This yielded a stable
structure of three major dimensions (since 1970, the structure is stable, as shown in the
1975, 1980 and 1985 studies). Each of these dimensions reflects a mainstream of ideolog-
ical thought in Western Europe since the early nineteenth century: traditional conset--
vatism, (socio-economic) liberalism and sociaiism the concepts measuring these
philosophies load consistently on the left-right dimension (liberalism and socialism) and
on the libertarian-authoritarian dimension (conservatism). (See Middendorp, 1978,
1991:72-6 and 87-95, and note c to Table 1.) On the concept of ‘ideology’; see Midden-
dorp (1978:102-8, 1991:59-61).
6. The causal models presented in Fig. 2 and 3 below are similar to those proposed by
Niemoller and Van der Eijk (19X&1, 1987) including the religion variable. Educational
level and age arc introduced as external variables. Class (subjectively measured-class
identification) and income are seen as dependent on those two variables, as well as
CEES P. MIDDENDORP 259
religion (an index for church attendance or rather ‘religious involvement’, with lowest
value for ‘not having been raised in a particular religion and not considering oneself to
belong to a religious denomination’ and highest value for ‘considering oneself to belong
to a religious denomination and attending church at least once a week’). For the
measurement of the ideological dimensions, see note to Table 1. Left-right self-identifi-
cation is measured on the basis of a five points scale (see Fig. 1). The models have been
tested applying LISREL-procedures (see Joreskog and S&born, 1983; Saris and
Stronkhorst, 1984).
7. Regarding the left-right vote, the ordering of the parties is: (from left to right): (1) CPN,
PSP, PPR (Communists, Pacifistic Socialists and ‘Christian’ Radicals, united in 1989 into
‘Green Left’). (2) PvdA (Labour Pary), (3) Democrats 66, (4) Christian Democrats, (5)
Liberals (VVD), (6) Religious fundamentalists (SGP, GPV). For the libertarian vote, the
order is: (1) CPN, PSP, PPR, (2) Democrats, (3) PvdA-Labour Party, (4) Liberals-WD,
(5) Christian Democrats, (6) Religious fundamentalists- SGP, GPV. Note that the order-
ings of the (groups of) parties is based on their mean position-m terms of supporters’
opinions-on the two ideological dimensions throughout the whole period of study
(197085). Some smaller parties (D66-Democrats, and PPR-Radicals) in particular
have changed their positions relative to those of other parties.
8. This contradictory effect of educational level on left-right self-identification-and
through that on both the left-right and the libertarian-authoritarian vote is expressed in
relatively low total effect on the vote. This phenomenon will be explored in future
research.
9. Analyses are based on the Dutch National Election Study 1982. The religious vote is very
similar to the libertarian-authoritarian vote-see e.g. Broos (1987).
10. See Middendorp (1991, ch. 9) for a general discussion on both left-right self-placement
and materialism-post-materialism.
References
A.J.C. Broos, ‘Religie en Stemgedrag: een aanpassing en uitbreiding van het model van
Niemoller en Van der Eijk’, Paper gepresenteerd op de workshop ‘Ideologie als deter-
minant van stemgedrag’, Nederlandse Kring voor Wetenschap der Politiek, Amersfoort,
1987.
J.W. van Deth, Political Science as no-risk policy’, Acta Politica, 2: 1986, pp. 185-9.
J.W. van Deth, P.A.T.M. Geurts, ‘Value orientation, left-right placement and voting’,
European Journal of Political Research, 17, 1989, pp. 17-34.
C. van der Eijk and B. Niemoller, Electoral Change in the Netherlands. Empirical results
and Methods of measurement, (Amsterdam: CT-Press, 1983).
S.C. Flanagan, ‘Changing values in advanced industrial societies’, Comparative Political
Studies, 1982, 14, pp. 403-44.
S.C. Flanagan, ‘Value change in industrial societies’, American Political Science Review,
1987, pp. 1303-18.
R. Inglehart, The Silent Revolution, (Princeton: Princeton University Press, 1977).
R. Inglehart, ‘From Class-based to Value-based Politics’, in: P. Mair (editor) The West
European Party System (Oxford, University Press, 1990), pp. 266-282.
R. Inglehart and J.-R. Rabier, ‘Political Realignment in Advanced Industrial Society: From
Class-based to Quality of Life Politics’, Government and Opposition, 21, 1986, pp. 456-79.
K.G. Joreskog and D. Sorbom, LISTEL Vl User’s Guide, Uppsale, Department of Statistics, 1983.
0. Knutsen, ‘Political Cleavages and Political Realignment in Norway. The New Political
Thesis Reexamined. Scandinavian Political Studies, 1986, 9, pp. 235-63.
0. Knutsen, ‘The impact of structural and ideological party cleavages in West-European
democracies; a comparative empirical analysis’, British Journal of Political Science, 1988,
18, pp. 324-52.
0. Knutsen, ‘The priorities of materialist and post-materialist values in the Nordic countries;
A Five-Nation Comparison; Scandinavian Political Studies, 1989, 12, 3, pp. 221-43.
260 Left-Right Set-iden~~ication